Hi David, Wow, thanks for having found this out. I had a look at the out files, and there is indeed a weird 20100422_04_control_07.c. 10165.10165.10.out file in there (see the attached tgz file for 20100422_04_control_07.c). I had a quick look in the other files and there is nothing strange like that in the other runs. I still do not quite understand how this could have happened as an output of Sequest, but that was definitively what was causing the bug obviously. Funny thing was that TPP v4.0 was not "reacting" to this weird out file the same way. Anyway, thanks again a lot for having figured this out. That was a tough one that took me weeks of my time and a lot of my nerves. And thanks for having fixed that in the next TPP release. Hope that this bug did not affect too many people's work. Best (and respectful) regards, Ludovic
On Jun 3, 12:52 am, David Shteynberg <[email protected]> wrote: > Hi Andreas, > > I suspect this problem was only with your sequest search because of > the offending spectrum I pointed out. I've corrected the source of > PeptideProphet so it will in the future simply spit out a warning and > ignore spectra where the encoded charge doesn't match assumed charge. > For now you can get your sequest analysis to go through correctly by > removing the offending spectrum. I need to debug the pepXML tool that > generated the file to solve the underlying converter issue. Do you > have the .out files available? > > Thanks, > -David > > On Wed, Jun 2, 2010 at 3:37 PM, Andreas Quandt <[email protected]> > wrote: > > hey david, > > there is another question, i would like to ask: when trying to trace the > > problem we also analyzed these mzxml files with other programs such as > > xtandem, mascot, and omssa. by looking at their results we did not oberserve > > any suspicious behavior. hence, do you think the problem you described does > > only occur when using sequest or is this a more general problem? > > cheers, > > andreas > > > On Thu, Jun 3, 2010 at 12:24 AM, Andreas Quandt <[email protected]> > > wrote: > > >> hey david, > >> many thanks for looking into it! > >> i was wondering if you could point me to a source where conditions such as > >> not using a '.' character in the file's basename are described because i > >> was > >> not aware of that. > >> many thanks in advance, > >> andreas > > >> On Wed, Jun 2, 2010 at 12:17 AM, David Shteynberg > >> <[email protected]> wrote: > > >>> Hello Ludovic, > > >>> I think the problem might be in the pepXML generated in your pipeline. > >>> The offending entry is in the file 20100422_04_control_07.c.pep.xml : > > >>> <spectrum_query spectrum="20100422_04_control_07.c.10165.10165.10" > >>> start_scan="10165" end_scan="10165" precursor_neutral_mass="3602.2759" > >>> assumed_charge="1" index="307" retention_time_sec="5514.3"> > > >>> Looks like the assumed charge opf 1+ doesn't match the encoded charge > >>> in the spectrum name (the number after the last dot "10" in this > >>> case). > > >>> If you simply delete this entry from the file the fvalues and > >>> probabilities should line up again. I will change the code to error > >>> out when this scenario is encountered in the future. > > >>> To address the underlying issue it looks like the conversion to pepXML > >>> failed on this entry for some reason. How were the pepXML files > >>> generated? Can you send me the original .out files so I can test the > >>> Out2XML sequest pepXML converter tool. This could also be caused by > >>> the '.' character in the basename of your files. Only alphanumeric > >>> and underscore characters are allowed there traditionally. > > >>> Thanks, > >>> -David > > >>> On Mon, May 31, 2010 at 3:37 PM, David Shteynberg > >>> <[email protected]> wrote: > >>> > Hello Ludovic, > > >>> > Yes, I see that there is a difference in the output files. I think > >>> > the problem is that the output of probabilities and fvals is > >>> > misaligned from the spectra in the pep.xml file. You'll see that the > >>> > next spectrum has the correct value. This definitely points to a bug > >>> > in the 4.3 version that is quite dated now. We are close to releasing > >>> > a new version. Are you able to test the latest SVN trunk version of > >>> > the software on your system to check if the bug is still present > >>> > there? > > >>> > Thanks, > >>> > -David > > >>> > On Mon, May 31, 2010 at 1:53 AM, lgillet <[email protected]> > >>> > wrote: > >>> >> Hi David, > > >>> >> No, I confirm again that I do find a difference upon running the > >>> >> xinteract command with the files in different orders (I confirm I also > >>> >> see those differences on TPP 4.3.1 installed on Unix and on > >>> >> WindowsXP). > >>> >> I have re-run the command on the same folder, on the same files, to > >>> >> avoid any confusion about file names or so. The result of the first > >>> >> command was named NormalOrder, and the second ScrambledOrder. > >>> >> I have made another zip file (xinteract-output.zip) with the output of > >>> >> my commands (such that you can check if I do anything wrong by running > >>> >> the command). Also at the end of the text file, I have copy-pasted > >>> >> some line from the summary visualisation from the web interface, with > >>> >> different filter criteria. > >>> >> I also attached in the zip the output of the 2 interact.pep.xml. > >>> >> If you look at a diff between both files, you will find plenty of > >>> >> differences on some spectra. For example, you can have a look at: > > >>> >> 20100422_01_control_04.c.07416.07416.3 => fval = 4.88 Vs. 0.058 (in > >>> >> NormalOrder Vs. ScrambledOrder resp.) > >>> >> 20100422_01_control_05.c.08545.08545.3 => fval = 5.36 Vs. completely > >>> >> absent (in NormalOrder Vs. ScrambledOrder resp.) > > >>> >> Finally, could you please run those 4 pep.xml files on your server on > >>> >> TPP 4.0 and TPP 4.3 by yourself? > >>> >> You may realize by yourself then that there are not only "some > >>> >> differeces" but the differences (especially if you look at the decoy > >>> >> protein hits) is terrible. > > >>> >> Thanks and let me know if something is still not clear. > > >>> >> Ludovic > > >>> >> On May 27, 7:33 pm, David Shteynberg <[email protected]> > >>> >> wrote: > >>> >>> Hi Ludovic, > > >>> >>> It is completely normal to expect some difference in the results > >>> >>> between version of the software since the models maybe slightly > >>> >>> different in new a version due to optimization, bug fixes and the > >>> >>> sort. Hopefully the new analysis is able to increase your correct > >>> >>> identifications at a set error rate. > > >>> >>> When I run your data through our 4.3.1 pipeline I get your result in > >>> >>> the scrambled analysis (regardless of the order in which I specify my > >>> >>> input pepxml files). The difference in *your* two analyses is due to > >>> >>> the difference in your input files. Here is the relevant info from > >>> >>> your two 4.3.1 analyses: > > >>> >>> interact-TPP-V4.3.pep.xml has 8931 spectra in charge 2+ that it > >>> >>> models: > > >>> >>> <mixture_model precursor_ion_charge="2" comments="using no. tolerable > >>> >>> trypsin term. [ntt] 0 data as pseudonegatives" > >>> >>> prior_probability="0.427" est_tot_correct="3830.1" > >>> >>> tot_num_spectra="8931" num_iterations="28"> > > >>> >>> interact_TPP-V4.3_scrambled.pep.xml has 8929 spectra in charge 2+ > >>> >>> that > >>> >>> it models: > > >>> >>> <mixture_model precursor_ion_charge="2" comments="using no. tolerable > >>> >>> trypsin term. [ntt] 0 data as pseudonegatives" > >>> >>> prior_probability="0.427" est_tot_correct="3829.1" > >>> >>> tot_num_spectra="8929" num_iterations="28"> > > >>> >>> Since the inputs are different in the two analyses the results will > >>> >>> be > >>> >>> different. Please verify that the inputs your are giving to the two > >>> >>> analyses in different order are *not identical*. Can you verify > >>> >>> this? > > >>> >>> Thanks, > >>> >>> -David > > >>> >>> On Thu, May 27, 2010 at 8:12 AM, lgillet <[email protected]> > >>> >>> wrote: > >>> >>> > Hi David, > > >>> >>> > TPP is installed in different servers in our Institute. I have re- > >>> >>> > uploaded a new file (lgillet_interact-again.zip) for which the TPP > >>> >>> > xinteract was performed on the same server and with different > >>> >>> > versions > >>> >>> > of the TPP. You can see that the results are still very different, > >>> >>> > even the scrambled case. > >>> >>> > Note that I used the version TPP v4.3 JETSTREAM rev 1, Build > >>> >>> > 201004201202 (linux); which I do not know if it is the same as the > >>> >>> > SVN > >>> >>> > TPP that you mentioned or a "nightly-built". > >>> >>> > - Can you re-confirm my results using your installation of TPP with > >>> >>> > my > >>> >>> > 4 pep.xml files? > >>> >>> > - Can you re-confirm the differences in decoy % using your > >>> >>> > installation of TPP between TPP V4.0 and V4.3 with my 4 pep.xml > >>> >>> > files? > >>> >>> > Thanks again, > >>> >>> > Ludovic > > >>> >>> > On May 26, 9:01 pm, David Shteynberg > >>> >>> > <[email protected]> > >>> >>> > wrote: > >>> >>> >> Hi Ludovic, > > >>> >>> >> I was unable to duplicate the different results on different order > >>> >>> >> of > >>> >>> >> input using the latest version of SVN tpp or version 4.3.1. I > >>> >>> >> noticed > >>> >>> >> that your two analyses point to different locations. Are you sure > >>> >>> >> that the files at these locations are identical? > > >>> >>> >> Thanks, > >>> >>> >> -David > > >>> >>> >> On Wed, May 26, 2010 at 10:47 AM, lgillet > >>> >>> >> <[email protected]> wrote: > >>> >>> >> > Hi David, > > >>> >>> >> > all my apologizes, the rar file got corrupted probably during > >>> >>> >> > the > >>> >>> >> > upload (the original on my HD was fine). > >>> >>> >> > I have uploaded again a zip file this time: > >>> >>> >> > lgillet_pepxml-again2.zip > >>> >>> >> > I hope that works this time (after download, I can decompress it > >>> >>> >> > back). > >>> >>> >> > Thanks for having a look at this issue. > >>> >>> >> > Best, > >>> >>> >> > Ludovic > > >>> >>> >> > On May 25, 7:24 pm, David Shteynberg > >>> >>> >> > <[email protected]> > >>> >>> >> > wrote: > >>> >>> >> >> Hi Ludovic, > > >>> >>> >> >> It seems the file you uploaded lgillet_pepxml_for_TPP4.3.rar is > >>> >>> >> >> corrupted. At least I am unable to open it. Please upload > >>> >>> >> >> again. > > >>> >>> >> >> Thanks, > >>> >>> >> >> -David > > >>> >>> >> >> On Wed, May 19, 2010 at 2:54 AM, lgillet > >>> >>> >> >> <[email protected]> wrote: > >>> >>> >> >> > Hi David, Hi Natalie, > > >>> >>> >> >> > I just posted the 4 pepxml files which give me the most > >>> >>> >> >> > striking > >>> >>> >> >> > differences in results between TPP-V4.0 and TPP-V4.3: > >>> >>> >> >> > lgillet_pepxml_for_TPP4.3.rar. I also posted the results > >>> >>> >> >> > (interact.pep.xml) which I obtain from running TPP-V4.0, > > ... > > read more » -- You received this message because you are subscribed to the Google Groups "spctools-discuss" group. To post to this group, send email to [email protected]. To unsubscribe from this group, send email to [email protected]. For more options, visit this group at http://groups.google.com/group/spctools-discuss?hl=en.
